Huffman and linear scanning methods with statistical language models

Area of Science:

• Augmentative and Alternative Communication (AAC)
• Human-Computer Interaction
• Assistive Technology

Background:

• Current scanning methods for Augmentative and Alternative Communication (AAC) devices, primarily row/column scanning, offer limited efficiency for text generation.
• There is a need for improved scanning techniques to enhance communication speed and usability for individuals relying on AAC devices.

Purpose of the Study:

• To introduce and evaluate Huffman scanning, a novel method integrating statistical language models with binary-switch, static-grid AAC interfaces.
• To compare the typing performance of Huffman scanning against traditional row/column scanning methods.

Main Methods:

• Developed and implemented Huffman scanning, a new method for AAC text generation using statistical language models.
• Conducted comparative typing tests involving 16 participants without disabilities and one individual with locked-in syndrome.
• Assessed typing speeds across different scanning methods and scan rates, including individually calibrated rates.

Main Results:

• Huffman scanning demonstrated significant typing speed improvements for participants without disabilities compared to all tested row/column scanning methods.
• The individual with locked-in syndrome also exhibited enhanced typing speeds with Huffman scanning.
• Faster typing speeds were achieved with Huffman scanning at a more relaxed scan rate than faster, individually calibrated rates.

Conclusions:

• Huffman scanning presents a promising and faster alternative to conventional row/column scanning for AAC text generation.
• The integration of statistical language models within Huffman scanning enhances usability and efficiency for diverse AAC users.
• Further research into optimized scan rates for Huffman scanning could yield even greater performance gains.